Ontology-Based ML Service Composition for Self-Adaptive Deployment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning applications require significant programming knowledge, lack interoperability across different contexts and data schemas, and struggle with self-adaptive architectures that can adjust to changes in online data metrics or QoS violations, leading to inefficiencies in model performance and deployment.
Innovation Solution
A machine learning platform that generates a library of components, uses a chatbot for intuitive interface, and employs self-adjusting features to create client-agnostic models, incorporating adaptive pipelining and ontology-based services for high-precision searches, allowing minimal human intervention and automatic model updates based on performance metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing machine learning tools are used to generate machine learning applications, then model performance can be optimized, but significant programming knowledge is required and the process becomes complex
Solution Approach 1:
The patent introduces an intermediary system that acts as a mediator between the user and machine learning model generation. This system automatically generates machine learning models by interfacing with existing tools and services, shielding users from programming complexity while maintaining model performance optimization capabilities
Solution Approach 2:
The system enables self-service by automatically generating machine learning models without requiring user programming knowledge. The automated model generation process services the user's needs directly, eliminating the need for users to manually program complex machine learning applications
2Adaptability or versatility
If existing machine learning models are used, then specific problems can be solved, but interoperability across different contexts and data schemas is lacking
Solution Approach 1:
The patent implements universality by designing a system that can handle multiple data schemas and contexts through a unified interface. The automated model generation process creates models that are adaptable to different contexts while maintaining reliable interoperability across various data formats and schemas
3Adaptability or versatility
If manual reconciliation processes are used for data schema matching, then data compatibility can be achieved, but the process becomes time-consuming and tedious
Solution Approach 1:
The system performs preliminary action by automatically preparing and reconciling data schemas before model generation. The automated process pre-processes data compatibility issues, eliminating the need for time-consuming manual reconciliation operations
Solution Approach 2:
The patent replaces the mechanical manual reconciliation process with an automated computational system. The automated schema matching and data compatibility processes substitute for tedious manual operations, achieving the same data compatibility goals without time loss
4Adaptability or versatility
If machine learning applications are designed with fixed architectures, then development is simpler, but self-adaptive capabilities to adjust to changes in online data metrics are lost
Solution Approach 1:
The patent implements dynamics by creating machine learning models with adaptive architectures that can change in response to online data metrics. The system dynamically adjusts model parameters and structures based on performance feedback, enabling self-adaptation while managing architecture complexity through automated control mechanisms
5Reliability
If existing tools focus on maximizing classification accuracy, then model performance improves, but the ability to adapt at run-time due to QoS violations is reduced
Solution Approach 1:
The patent implements feedback mechanisms that monitor both classification accuracy and QoS metrics in real-time. The system uses this feedback to dynamically adjust model behavior and architecture, maintaining high accuracy while enabling run-time adaptation to QoS violations through continuous performance monitoring and automated response
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A server system may receive a plurality of inputs that identifies a location of data, describes a prediction for the machine learning application, and one or more constraints for the machine learning application. The server system may access a memory containing one or more ontologies of the data. The server system may extract one of more attributes of the data sets to find a first ontology that correlates to the prediction for the machine learning application according to the one or more constraints. The server system may compose a product graph based on the on the first ontology, the one or more constraints, and one or more previous product graphs stored in the memory, wherein the product graph relates the one or more data objects to a collection of nodes and edges, wherein the edges represent links between the nodes comprising a basic unit of a data structure.